A method and system for dynamic optimization of air conditioning based on deep learning

Abstract

The present invention relates to a deep learning-based dynamic air conditioning optimization method and system, belonging to the technical field of intelligent air conditioning control. The method comprises acquiring air conditioning operation data, constructing an air conditioning operation dataset based on the air conditioning operation data through air conditioning data preprocessing, obtaining an air conditioning user preference model based on the air conditioning operation dataset through multi-layer perceptron training and testing, and constructing an air conditioning user preference DQN model based on the air conditioning user preference model by initializing a DQN network; collecting environmental sensor data in real time, outputting a user air conditioning optimal control strategy based on the environmental sensor data through the air conditioning user preference DQN model, and controlling the air conditioning equipment in real time through the air conditioning control system based on the user air conditioning optimal control strategy, thereby dynamically optimizing air conditioning parameters based on user comfort and energy-saving goals.

Description

Technical Field

[0001] The present invention belongs to the technical field of air-conditioning intelligent control, and in particular relates to a method and system for dynamic optimization of air-conditioning based on deep learning. Background Art

[0002] Energy conservation, emission reduction, and intelligent control have become key trends in air conditioning system design. Traditional air conditioning control systems typically adjust temperature and humidity based on preset parameters, lacking dynamic response to actual environmental changes and user needs. This results in energy waste and unstable indoor comfort. Therefore, it is crucial to design a method and system that can dynamically adjust air conditioning operating parameters in real time based on environmental changes and user needs.

[0003] However, traditional machine learning algorithms suffer from low prediction accuracy when dealing with complex nonlinear problems in air conditioning systems. They are unable to accurately predict user needs and environmental changes, leading to inaccurate control parameter adjustments. Furthermore, traditional algorithms are typically trained on specific datasets and lack adaptability, which can result in the system being unable to adapt to environmental changes or shifts in user habits. Furthermore, traditional algorithms rely heavily on manual feature extraction and design, easily overlooking key information in sensor data, which in turn impacts the optimization of control strategies. Summary of the Invention

[0004] In order to solve the above problems existing in the prior art, the present invention proposes an air conditioning dynamic optimization method and system based on deep learning.

[0005] The purpose of the present invention can be achieved through the following technical solutions:

[0006] A deep learning-based air conditioning dynamic optimization method, comprising:

[0007] Acquire air conditioning operation data, and construct an air conditioning operation data set through air conditioning data preprocessing based on the air conditioning operation data, wherein the air conditioning operation data includes environmental sensor data, user behavior data, and air conditioning control parameters;

[0008] According to the air conditioning operation data set, a multi-layer perceptron training and testing is used to obtain an air conditioning user preference model;

[0009] Constructing an air-conditioning user preference DQN model by initializing a DQN network according to the air-conditioning user preference model;

[0010] Environmental sensor data is collected in real time, an optimal control strategy for user air conditioning is outputted through the air conditioning user preference DQN model according to the environmental sensor data, and air conditioning equipment is controlled in real time through the air conditioning control system according to the optimal control strategy for user air conditioning.

[0011] Preferably, the acquiring of historical air-conditioning operation data and constructing an air-conditioning operation data set through air-conditioning data preprocessing according to the air-conditioning operation data includes:

[0012] The environmental sensor data includes indoor temperature, outdoor temperature, indoor humidity, and outdoor humidity; the user behavior data includes air conditioning temperature setting data and air conditioning wind speed setting data; and the air conditioning control parameters include air conditioning start and stop parameters, air conditioning temperature adjustment parameters, and air conditioning wind speed adjustment parameters;

[0013] The air conditioning data preprocessing is to obtain air conditioning operation standard data by normalizing the air conditioning operation data by means;

[0014] An air conditioning operation data set is constructed according to the air conditioning operation standard data.

[0015] Preferably, the air-conditioning user preference model obtained by multi-layer perceptron training and testing according to the air-conditioning operation data set includes:

[0016] A multilayer perceptron is obtained by selecting multilayer perceptron nodes according to the air-conditioning operation data set, wherein the multilayer perceptron nodes include the number of input layer nodes, the number of hidden layer nodes, and the number of output layer nodes;

[0017] According to the air-conditioning operation data set, an air-conditioning operation training set and an air-conditioning operation test set are obtained by dividing the data set;

[0018] Calculating the multilayer perceptron loss by forward propagation according to the air conditioning operation training set, wherein the multilayer perceptron loss is calculated by a mean square error loss function;

[0019] Calculating gradients according to the multilayer perceptron and updating weights and biases of the multilayer perceptron, and iteratively training until the multilayer perceptron reaches a preset number of iterations;

[0020] The air-conditioning user preference model is obtained by evaluating the multi-layer perceptron through the root mean square error according to the air-conditioning operation test set.

[0021] Preferably, constructing the air-conditioning user preference DQN model by initializing the DQN network according to the air-conditioning user preference model includes:

[0022] Constructing an air conditioning state space, wherein the air conditioning state space includes environmental sensor data and user behavior data;

[0023] Constructing an air conditioning action space, wherein the air conditioning action space includes increasing temperature, decreasing temperature, maintaining temperature, increasing wind speed, decreasing wind speed, and maintaining wind speed;

[0024] Encoding the air conditioning action space by air conditioning action parameterization, wherein the air conditioning action parameterization includes increasing temperature as 0, decreasing temperature as 1, maintaining temperature as 2, increasing wind speed as 3, decreasing wind speed as 4, and maintaining wind speed as 5;

[0025] The total air conditioning reward value is calculated by designing an air conditioning reward mechanism, and the total air conditioning reward value is the sum of the user comfort reward value and the energy saving reward value;

[0026] Outputting an air-conditioning action A through the ε-greedy strategy according to the current environmental state H, wherein the air-conditioning action A is an air-conditioning action in the air-conditioning action space;

[0027] Air-conditioning experience training data is obtained by randomly sampling the air-conditioning experience replay according to the air-conditioning experience replay buffer;

[0028] The air-conditioning user preference DQN model is obtained by training and testing the DQN network through a deep neural network approximation Q function according to the air-conditioning experience tuple.

[0029] Preferably, the air conditioning reward mechanism includes:

[0030] Preset user comfort range and air conditioning energy consumption benchmark value, the user comfort range includes temperature comfort range and wind speed comfort range;

[0031] Calculate the user comfort reward value based on the air conditioning operation, and determine if the temperature and wind speed in the next air conditioning state are within the user comfort range. If yes, add 1 to the user comfort reward value; otherwise, subtract 1 from the user comfort reward value.

[0032] The energy-saving bonus value is calculated based on the air-conditioning operation, and it is determined whether the air-conditioning energy consumption is lower than the air-conditioning energy consumption baseline value after the next air-conditioning action is executed. If yes, the energy-saving bonus value is increased by 0.5, otherwise, the energy-saving bonus value is subtracted by 0.5.

[0033] Preferably, outputting the air-conditioning action A through the ε-greedy strategy according to the current environmental state H includes:

[0034] A preset exploration rate ε value is the probability of randomly selecting an action;

[0035] Randomly generate an air conditioning exploration random number in the interval [0, 1) according to the exploration rate ε value;

[0036] When the air conditioning exploration random number is less than the exploration rate ε value, an air conditioning action A is randomly selected from the air conditioning action space;

[0037] When the air-conditioning action space is greater than or equal to the exploration rate ε value, the Q values ​​of all possible actions under the current environment state H are predicted through the DQN network, and the air-conditioning action A with the highest Q value is selected.

[0038] Preferably, the air-conditioning user preference DQN model is obtained by training and testing the DQN network through a deep neural network approximation Q function according to the air-conditioning experience tuple, including:

[0039] Obtaining a DQN initial network by randomly initializing network weights and biases according to the DQN network;

[0040] Calculating the air conditioning action loss using a mean square error loss function based on the DQN initial network, where the air conditioning action loss is the difference between the predicted Q value of the air conditioning action and the target Q value of the air conditioning action;

[0041] The network weights are updated by the Adam optimizer until the DQN initial network converges, thereby obtaining the air conditioner user preference DQN model.

[0042] An air conditioning dynamic optimization system based on deep learning that applies the air conditioning dynamic optimization method based on deep learning, comprising an air conditioning data acquisition and preprocessing module, an air conditioning user preference module, an air conditioning dynamic optimization module, and an air conditioning control module;

[0043] The air conditioning data acquisition and preprocessing module is used to collect air conditioning operation data and normalize the mean to obtain air conditioning operation standard data, and then construct an air conditioning operation data set;

[0044] The air-conditioning user preference module is used to obtain an air-conditioning user preference model by training a multi-layer perceptron based on the air-conditioning operation data set to identify the user's air-conditioning preference;

[0045] The air-conditioning dynamic optimization module is used to train the DQN network based on the air-conditioning user preference model through air-conditioning experience replay and ε-greedy strategy to obtain the air-conditioning user preference DQN model;

[0046] The air-conditioning control module is used to output the user air-conditioning optimal control strategy through the air-conditioning user preference DQN model according to the current environmental sensor data, and to adjust the air-conditioning equipment in real time through the air-conditioning control system according to the user air-conditioning optimal control strategy to achieve dynamic optimization of air-conditioning parameters based on user comfort and energy-saving goals.

[0047] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the method for dynamic optimization of air conditioning based on deep learning is implemented.

[0048] A storage medium containing computer-executable instructions, which, when executed by a computer processor, are used to perform the above-mentioned deep learning-based air conditioning dynamic optimization method.

[0049] The beneficial effects of the present invention are:

[0050] (1) The air conditioning user preference DQN model is used to achieve adaptive dynamic optimization of the air conditioning. The air conditioning operation is intelligently adjusted according to the current environmental status and user air conditioning preferences, effectively reducing unnecessary energy consumption and achieving energy saving and consumption reduction. At the same time, it better meets the personalized needs of users, thereby improving user comfort.

[0051] (2) The air-conditioning user preference model is obtained through multi-layer perceptron training and testing, which can learn the user's air-conditioning preferences, provide optimization targets for subsequent air-conditioning dynamic optimization, and significantly improve user comfort.

[0052] (3) By designing an air conditioning reward mechanism, energy-saving goals and user air conditioning preferences are taken into consideration at the same time, and dynamic optimization of air conditioning is achieved by balancing air conditioning energy consumption and user needs.

[0053] (4) Through the air-conditioning experience replay mechanism and the ε-greedy strategy, the model achieves a balance between exploring new strategies and utilizing known information, which improves the stability and efficiency of the model optimization control strategy process and ensures that the air-conditioning user preference DQN model can quickly respond to dynamic changes in the environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] To facilitate understanding by those skilled in the art, the present invention is further described below with reference to the accompanying drawings.

[0055] Figure 1 This is a flow chart of a deep learning-based air conditioning dynamic optimization method of the present invention. DETAILED DESCRIPTION

[0056] In order to further illustrate the technical means and effects adopted by the present invention to achieve the predetermined purpose of the invention, the specific implementation methods, structures, features and effects of the present invention are described in detail below in conjunction with the accompanying drawings and preferred embodiments.

[0057] See also Figure 1 , an air conditioning dynamic optimization method based on deep learning, comprising:

[0058] S1: Acquire air conditioning operation data, and construct an air conditioning operation data set through air conditioning data preprocessing based on the air conditioning operation data, wherein the air conditioning operation data includes environmental sensor data, user behavior data, and air conditioning control parameters;

[0059] S2: obtaining an air-conditioning user preference model through multi-layer perceptron training and testing based on the air-conditioning operation data set;

[0060] S3: Constructing an air-conditioning user preference DQN model by initializing a DQN network according to the air-conditioning user preference model;

[0061] S4: Collect environmental sensor data in real time, output a user air-conditioning optimal control strategy through the air-conditioning user preference DQN model according to the environmental sensor data, and adjust the air-conditioning equipment in real time through the air-conditioning control system according to the user air-conditioning optimal control strategy.

[0062] In this embodiment, acquiring historical air-conditioning operation data and constructing an air-conditioning operation data set through air-conditioning data preprocessing according to the air-conditioning operation data specifically includes:

[0063] The environmental sensor data includes indoor temperature, outdoor temperature, indoor humidity, and outdoor humidity; the user behavior data includes air conditioning temperature setting data and air conditioning wind speed setting data; and the air conditioning control parameters include air conditioning start and stop parameters, air conditioning temperature adjustment parameters, and air conditioning wind speed adjustment parameters;

[0064] The air conditioning data preprocessing is to obtain air conditioning operation standard data by normalizing the air conditioning operation data by means;

[0065] An air conditioning operation data set is constructed according to the air conditioning operation standard data.

[0066] Specifically, the data units of the indoor temperature, the outdoor temperature and the air-conditioning temperature setting data are uniformly Celsius, and the data units of the indoor humidity and the outdoor humidity are uniformly percentage.

[0067] In this embodiment, the air-conditioning user preference model is obtained by multi-layer perceptron training and testing based on the air-conditioning operation data set, specifically by the following steps:

[0068] S201: obtaining a multilayer perceptron by selecting multilayer perceptron nodes according to the air-conditioning operation data set, wherein the multilayer perceptron nodes include the number of input layer nodes, the number of hidden layer nodes, and the number of output layer nodes;

[0069] S202: obtaining an air-conditioning operation training set and an air-conditioning operation test set by dividing the air-conditioning operation data set into a data set according to a ratio of 7:3;

[0070] S203: Calculating the multilayer perceptron loss by forward propagation according to the air-conditioning operation training set, wherein the multilayer perceptron loss is calculated by a mean square error loss function;

[0071] S204: Calculating gradients according to the multilayer perceptron and updating weights and biases of the multilayer perceptron, and iteratively training until the multilayer perceptron reaches a preset number of iterations;

[0072] S205: Evaluating the multilayer perceptron by root mean square error according to the air-conditioning operation test set to obtain an air-conditioning user preference model.

[0073] In this embodiment, the number of hidden layer nodes is determined by cross-validation, the number of input layer nodes is 4 nodes, the number of output layer nodes is 2 nodes, the difference between the predicted value of the air-conditioning operation data and the actual value of the air-conditioning operation data is measured by the mean square error loss function, the mean square error loss function is minimized by the Adam optimizer, and the model expression ability is maintained by the Relu activation function.

[0074] In this embodiment, the air conditioner user preference DQN model is constructed by initializing the DQN network according to the air conditioner user preference model, specifically by the following steps:

[0075] S301: Constructing an air-conditioning state space, wherein the air-conditioning state space includes environmental sensor data and user behavior data;

[0076] S302: Constructing an air conditioning action space, wherein the air conditioning action space includes increasing temperature, decreasing temperature, maintaining temperature, increasing wind speed, decreasing wind speed, and maintaining wind speed;

[0077] S303: Air conditioning action encoding is performed on the air conditioning action space through air conditioning action parameterization, wherein the air conditioning action parameterization includes increasing temperature as 0, decreasing temperature as 1, maintaining temperature as 2, increasing wind speed as 3, decreasing wind speed as 4, and maintaining wind speed as 5.

[0078] In this embodiment, the step size of temperature adjustment is 1 degree Celsius, and the step size of wind speed adjustment is 1 wind speed level. The air-conditioning state space is a set of air-conditioning state variables that affect the operation of the air-conditioning system. Each state in the air-conditioning state space can be represented as a multidimensional vector, where each dimension corresponds to a state variable. The air-conditioning action space is the control operation that the air-conditioning can perform. The actions in the air-conditioning action space are converted into a form that can be understood and executed by the model through action parameterization.

[0079] S304: Calculating a total air conditioning reward value by designing an air conditioning reward mechanism, wherein the total air conditioning reward value is the sum of the user comfort reward value and the energy-saving reward value;

[0080] The air conditioning reward mechanism includes:

[0081] S304-1: Preset a user comfort range and an air conditioning energy consumption benchmark value, wherein the user comfort range includes a temperature comfort range and a wind speed comfort range;

[0082] S304-2: Calculate the user comfort reward value based on the air conditioning operation, and determine whether the temperature and wind speed in the next air conditioning state are within the user comfort range. If yes, add 1 to the user comfort reward value; otherwise, subtract 1 from the user comfort reward value.

[0083] S304-3: Calculate the energy-saving bonus value based on the air-conditioning operation, and determine whether the air-conditioning energy consumption is lower than the air-conditioning energy consumption baseline value after the next air-conditioning action is executed. If yes, add 0.5 to the energy-saving bonus value; otherwise, subtract 0.5 from the energy-saving bonus value.

[0084] S305: Outputting an air-conditioning action A using an ε-greedy strategy according to the current environmental state H, where the air-conditioning action A is an air-conditioning action in the air-conditioning action space;

[0085] S305-1: Preset an exploration rate ε value, where the exploration rate ε value is the probability of randomly selecting an action;

[0086] S305-2: Randomly generate an air conditioning exploration random number in the interval [0, 1) according to the exploration rate ε value;

[0087] S305-3: When the air-conditioning exploration random number is less than the exploration rate ε value, an air-conditioning action A is randomly selected from the air-conditioning action space.

[0088] When the air-conditioning action space is greater than or equal to the exploration rate ε value, the Q values ​​of all possible actions under the current environment state H are predicted through the DQN network, and the air-conditioning action A with the highest Q value is selected.

[0089] S306: randomly extracting air-conditioning experience training data by replaying the air-conditioning experience according to the air-conditioning experience replay buffer;

[0090] Construct an air conditioning experience tuple (H, A, R, H') and store it in the air conditioning experience playback buffer, where H is the current environment state, A is the air conditioning action, R is the total air conditioning reward value, and H' is the new air conditioning state;

[0091] When the air-conditioning experience playback buffer is larger than the preset air-conditioning experience playback buffer storage value, the new air-conditioning experience overwrites the old air-conditioning experience.

[0092] S307: training and testing the DQN network using a deep neural network approximation Q function according to the air conditioning experience tuple to obtain an air conditioning user preference DQN model;

[0093] S307-1: Obtaining a DQN initial network by randomly initializing network weights and biases according to the DQN network;

[0094] S307-2: Calculating the air conditioning action loss using a mean square error loss function based on the DQN initial network, where the air conditioning action loss is the difference between the predicted Q value of the air conditioning action and the target Q value of the air conditioning action;

[0095] S307-3: Update the network weights using the Adam optimizer until the DQN initial network converges, thereby obtaining the air conditioner user preference DQN model.

[0096] An air conditioning dynamic optimization system based on deep learning includes an air conditioning data acquisition and preprocessing module, an air conditioning user preference module, an air conditioning dynamic optimization module, and an air conditioning control module;

[0097] The air conditioning data acquisition and preprocessing module is used to collect air conditioning operation data and normalize the mean to obtain air conditioning operation standard data, and then construct an air conditioning operation data set;

[0098] The air-conditioning user preference module is used to obtain an air-conditioning user preference model by training a multi-layer perceptron based on the air-conditioning operation data set to identify the user's air-conditioning preference;

[0099] The air-conditioning dynamic optimization module is used to obtain the air-conditioning user preference DQN model by training the DQN network through air-conditioning experience replay and ε-greedy strategy based on the air-conditioning user preference model;

[0100] The air-conditioning control module is used to output the user air-conditioning optimal control strategy through the air-conditioning user preference DQN model according to the current environmental sensor data, and to adjust the air-conditioning equipment in real time through the air-conditioning control system according to the user air-conditioning optimal control strategy to achieve dynamic optimization of air-conditioning parameters based on user comfort and energy-saving goals.

[0101] The computer storage medium of the embodiment of the present invention may adopt any combination of one or more computer-readable media. The computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or component, or any combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In this document, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device.

[0102] A computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, which carries computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium that can transmit, propagate, or transport a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0103] The program code included in the computer-readable medium can be transmitted with any appropriate medium, including but not limited to wireless, electric wire, optical cable, RF or the like, or any suitable combination thereof. The computer program code for performing the operation of the present invention can be written in one or more programming languages ​​or a combination thereof, and the programming language includes an object-oriented programming language such as Java, Smalltalk, C++, and also includes a conventional procedural programming language such as "C" language or similar programming language. The program code can be executed completely on the user's computer, partially on the user's computer, as an independent software package, partially on the user's computer and partially on a remote computer, or completely on a remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computer (for example, utilizing an Internet service provider to connect through the Internet).

[0104] The above description is merely a preferred embodiment of the present invention and does not constitute any form of limitation to the present invention. Although the present invention has been disclosed as a preferred embodiment as above, it is not intended to limit the present invention. Any person skilled in the art can make some changes or modifications to equivalent embodiments using the technical contents disclosed above without departing from the scope of the technical solution of the present invention. However, any simple modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention are still within the scope of the technical solution of the present invention.

Claims

1. A method for dynamic optimization of air conditioning based on deep learning, characterized in that: include: Acquire air conditioning operation data, and construct an air conditioning operation data set through air conditioning data preprocessing based on the air conditioning operation data, wherein the air conditioning operation data includes environmental sensor data, user behavior data, and air conditioning control parameters; According to the air conditioning operation data set, a multi-layer perceptron training and testing is used to obtain an air conditioning user preference model; Constructing an air-conditioning user preference DQN model by initializing a DQN network according to the air-conditioning user preference model; Collecting environmental sensor data in real time, outputting a user air conditioning optimal control strategy through the air conditioning user preference DQN model based on the environmental sensor data, and regulating the air conditioning equipment in real time through the air conditioning control system based on the user air conditioning optimal control strategy; The constructing of the air-conditioning user preference DQN model by initializing the DQN network according to the air-conditioning user preference model includes: Constructing an air conditioning state space, wherein the air conditioning state space includes environmental sensor data and user behavior data; Constructing an air conditioning action space, wherein the air conditioning action space includes increasing temperature, decreasing temperature, maintaining temperature, increasing wind speed, decreasing wind speed, and maintaining wind speed; Encoding the air conditioning action space by air conditioning action parameterization, wherein the air conditioning action parameterization includes increasing temperature as 0, decreasing temperature as 1, maintaining temperature as 2, increasing wind speed as 3, decreasing wind speed as 4, and maintaining wind speed as 5; The total air conditioning reward value R is calculated by designing an air conditioning reward mechanism, and the total air conditioning reward value is the sum of the user comfort reward value and the energy saving reward value; Outputting an air-conditioning action A through the ε-greedy strategy according to the current environmental state H, wherein the air-conditioning action A is an air-conditioning action in the air-conditioning action space; Air-conditioning experience training data is obtained by randomly sampling the air-conditioning experience replay according to the air-conditioning experience replay buffer; Construct an air conditioning experience tuple (H, A, R, H') and store it in the air conditioning experience playback buffer, where H is the current environment state, A is the air conditioning action, R is the total air conditioning reward value, and H' is the new air conditioning state; When the air-conditioning experience playback buffer is larger than the preset air-conditioning experience playback buffer storage value, the new air-conditioning experience overwrites the old air-conditioning experience; The air-conditioning user preference DQN model is obtained by training and testing the DQN network through a deep neural network approximation Q function according to the air-conditioning experience tuple.

2. The air conditioning dynamic optimization method based on deep learning according to claim 1, characterized in that: The acquiring of the air-conditioning operation data and constructing the air-conditioning operation data set through air-conditioning data preprocessing according to the air-conditioning operation data include: The environmental sensor data includes indoor temperature, outdoor temperature, indoor humidity, and outdoor humidity; the user behavior data includes air conditioning temperature setting data and air conditioning wind speed setting data; and the air conditioning control parameters include air conditioning start and stop parameters, air conditioning temperature adjustment parameters, and air conditioning wind speed adjustment parameters; The air conditioning data preprocessing is to obtain air conditioning operation standard data by normalizing the air conditioning operation data by means; An air conditioning operation data set is constructed according to the air conditioning operation standard data.

3. The air conditioning dynamic optimization method based on deep learning according to claim 1, characterized in that: The air-conditioning user preference model obtained by multi-layer perceptron training and testing according to the air-conditioning operation data set includes: A multilayer perceptron is obtained by selecting multilayer perceptron nodes according to the air-conditioning operation data set, wherein the multilayer perceptron nodes include the number of input layer nodes, the number of hidden layer nodes, and the number of output layer nodes; According to the air-conditioning operation data set, an air-conditioning operation training set and an air-conditioning operation test set are obtained by dividing the data set; Calculating the multilayer perceptron loss by forward propagation according to the air conditioning operation training set, wherein the multilayer perceptron loss is calculated by a mean square error loss function; Calculating gradients according to the multilayer perceptron and updating weights and biases of the multilayer perceptron, and iteratively training until the multilayer perceptron reaches a preset number of iterations; The air-conditioning user preference model is obtained by evaluating the multi-layer perceptron through the root mean square error according to the air-conditioning operation test set.

4. The air conditioning dynamic optimization method based on deep learning according to claim 1, characterized in that: The air conditioning reward mechanism includes: Preset user comfort range and air conditioning energy consumption benchmark value, the user comfort range includes temperature comfort range and wind speed comfort range; Calculate the user comfort reward value based on the air conditioning operation, and determine whether the temperature and wind speed in the next air conditioning state are within the user comfort range. If yes, add 1 to the user comfort reward value; otherwise, subtract 1 from the user comfort reward value; The energy-saving bonus value is calculated based on the air-conditioning operation, and it is determined whether the air-conditioning energy consumption is lower than the air-conditioning energy consumption baseline value after the next air-conditioning action is executed. If yes, the energy-saving bonus value is increased by 0.5, otherwise, the energy-saving bonus value is subtracted by 0.

5.

5. The air conditioning dynamic optimization method based on deep learning according to claim 1, characterized in that: Outputting the air conditioning action A according to the current environment state H through the ε-greedy strategy includes: A preset exploration rate ε value is the probability of randomly selecting an action; Randomly generate an air conditioning exploration random number in the interval [0, 1) according to the exploration rate ε value; When the air conditioning exploration random number is less than the exploration rate ε value, an air conditioning action A is randomly selected from the air conditioning action space; When the air-conditioning action space is greater than or equal to the exploration rate ε value, the Q values ​​of all possible actions under the current environment state H are predicted through the DQN network, and the air-conditioning action A with the highest Q value is selected.

6. The air conditioning dynamic optimization method based on deep learning according to claim 1, characterized in that: The air-conditioning user preference DQN model is obtained by training and testing the DQN network through a deep neural network approximating a Q function according to the air-conditioning experience tuple, including: Obtaining a DQN initial network by randomly initializing network weights and biases according to the DQN network; Calculating the air conditioning action loss using a mean square error loss function based on the DQN initial network, where the air conditioning action loss is the difference between the predicted Q value of the air conditioning action and the target Q value of the air conditioning action; The network weights are updated by the Adam optimizer until the DQN initial network converges, thereby obtaining the air conditioner user preference DQN model.

7. An air conditioning dynamic optimization system based on deep learning, applying the air conditioning dynamic optimization method based on deep learning according to claim 1, comprising an air conditioning data acquisition and preprocessing module, an air conditioning user preference module, an air conditioning dynamic optimization module, and an air conditioning control module, characterized in that: The air conditioning data acquisition and preprocessing module is used to collect air conditioning operation data and normalize the mean to obtain air conditioning operation standard data, and then construct an air conditioning operation data set; The air-conditioning user preference module is used to obtain an air-conditioning user preference model by training a multi-layer perceptron based on the air-conditioning operation data set to identify the user's air-conditioning preference; The air-conditioning dynamic optimization module is used to train the DQN network based on the air-conditioning user preference model through air-conditioning experience replay and ε-greedy strategy to obtain the air-conditioning user preference DQN model; The air-conditioning control module is used to output the user air-conditioning optimal control strategy through the air-conditioning user preference DQN model according to the current environmental sensor data, and to adjust the air-conditioning equipment in real time through the air-conditioning control system according to the user air-conditioning optimal control strategy to achieve dynamic optimization of air-conditioning parameters based on user comfort and energy-saving goals.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the air conditioning dynamic optimization method based on deep learning as described in any one of claims 1 to 6 is implemented.

9. A storage medium containing computer-executable instructions, characterized in that: When executed by a computer processor, the computer executable instructions are used to execute the air conditioning dynamic optimization method based on deep learning as described in any one of claims 1 to 6.

Images